CO Adsorption-Driven Surface Segregation of Pd on Au/Pd BimetallicSurfaces: Role of Defects and Effect on CO OxidationHyun You Kim†,* and Graeme Henkelman*
Department of Chemistry, University of Texas at Austin, Austin, Texas 78712-0165, United States
*S Supporting Information
ABSTRACT: We use density functional theory (DFT) tostudy CO-adsorption-induced Pd surface segregation in Au/Pdbimetallic surfaces, dynamics of Pd!Au swapping, effect ofdefects on the swapping rate, CO-induced Pd clustering, andthe reaction mechanism of CO oxidation. The strong CO-philic nature of Pd atoms supplies a driving force for thepreferential surface segregation of Pd atoms and Pd clusterformation. Surface vacancies are found to dramaticallyaccelerate the rate of Pd!Au swapping. We find that Pdclusters consisting of at least four Pd atoms prefer to bind O2rather than CO. These clusters facilitate the rapid dissociation of O2 and supply reactive oxygen species for CO oxidation. Ourfindings suggest that geometric, electronic, and dynamic effects should be considered in the function of bimetallic alloys ornanoparticles whose components asymmetrically interact with reacting molecules.KEYWORDS: density functional theory, gold, palladium, heterogeneous catalysis, adsorption-induced segregation, CO oxidation
1. INTRODUCTIONThe catalytic activity of bimetallic alloys or nanoparticles (NPs)can be optimized by controlling structural factors, such as thealloying element and concentration.1,2 The ensemble (geo-metric) and ligand (electronic) effects have been shown tosystematically alter the catalytic activity of bimetallic catalysts.3,4
Computational methods such as density functional theory(DFT) can effectively aid the design of bimetallic catalysts atthe atomic scale.5!10 Computational approaches generallyassume that the thermodynamically most favorable structureof clean bimetallic catalysts is stable. Experiments, however,question the generality of this assumption. Somorjai and co-workers reported that the core and shell elements of Pd (core)@Rh (shell) NP are reversible under ambient reactionconditions.11,12 Using ambient-pressure X-ray photoelectronspectroscopy, they showed that as the Pd@Rh NPs supportedon oxidized silicon wafer oxidize CO with NO (2CO + 2NO!2CO2 + N2), Pd is enriched in the surface layers, leading to astructural rearrangement to the Rh@Pd reverse core!shellgeometry. The original Pd@Rh core!shell structure wasrecovered when CO was removed from the gas phase. Chenand co-workers reported such a CO-induced Pt segregation inTiO2-supported Pt!Au NPs, as well.13
In the case of bimetallic surfaces, the Goodman groupreported that Pd segregated to the surface layer as Auoverlayered-Pd(100) bimetallic alloys were exposed to COoxidation conditions.14,15 More Pd was segregated in thesurface layer as the CO partial pressure was increased. Thesurface segregated Pd atoms were found to catalyze COoxidation, and the authors suggested that contiguous Pd atomsin the surface layer provided the catalytically active site. Their
rationale was that contiguous Pd atoms bind and dissociate theO2 molecule supplying O atoms for CO oxidation.Since adsorption- or reaction-induced surface segregation of
a specific element in bimetallic alloys and NPs affects theirchemical properties, information on the reaction- oradsorption-induced surface segregation is important for catalystdesign.Interatomic swapping of core and shell elements reported in
Pd- or Pt-based bimetallic catalysts is presumably driven by anasymmetrically strong CO adsorption on Pd or Pt atoms.13!16
DFT results presented by Soto-Verdugo and Metiu showedthat CO prefers to bind on Pd rather than on Au in Au/Pdbimetallic alloys.16 The same trend was reported for Pt/Au NPsby Chen and co-workers.13 Although several previous reportson the CO-induced preferential surface segregation suggestedthat the strong CO binding on Pd or Pt induces the atomicswapping, detailed information on the swapping process is stillsketchy.Here, we demonstrate the dynamics of the CO-adsorption-
driven Pd!Au swapping, Pd surface segregation, that occurs inthe Pd!Au(111) and Pd!Au(100) bimetallic alloys. We foundthat the relatively strong CO binding on Pd, as compared withthat on Au, stabilizes the Pd!CO* in the surface layer andsupplies a driving force for Pd surface segregation. Surface Auvacancies accelerate the Pd!Au swapping, highlighting theessential role of defects on the swapping dynamics. The
Received: July 30, 2013Revised: September 26, 2013
morphology of reactive species and the role of the Pdconcentration on the CO oxidation mechanism are discussed.
2. COMPUTATIONAL METHODWe performed spin-polarized DFT calculations in a plane-wavebasis with the VASP code17 and the PBE GGA functional.18
Valence electrons were described by plane waves up to anenergy cutoff of 290 eV, and the core electrons were describedwithin the projector augmented wave framework.19 To studythe Pd!Au swapping mechanism, 4 ! 4 Au(111) and Au(100)slabs with four atomic layers and 20 Å of vacuum thicknesswere constructed. A single subsurface Au atom was substitutedwith a Pd atom to give an in-plane Pd concentration of 1/16 =6.25%. The top two surface layers were relaxed duringgeometric optimization. We used a 2 ! 2 ! 1 k-points gridsampling of the Brillouin zone for all calculations. Sensitivitytests show that our results are robust with respect to thecalculation and model parameters, including the choice ofoxygen pseudopotential, k-point grid, cutoff energy, and sizeand thickness of the slab (see Supporting Information Table S1for details). The energy of CO adsorption on the Pd!Au(100)calculated with a harder oxygen pseudopotential and an energycutoff of 500 eV was changed by only 0.07 eV.Final convergence criteria for the electronic wave function
and geometry were 10!4 eV and 0.01 eV/Å, respectively. Weused the Gaussian smearing method with a width of 0.2 eV toimprove convergence with respect to states near the Fermilevel. The location and energy of transition states (TSs) werecalculated with the climbing-image nudged elastic bandmethod20,21
3. RESULTS AND DISCUSSION3-1. Energetics of Pd!Au Swapping in Clean Surfaces.
In vacuum, a Au-covered Pd overlayer is the thermodynamicallymost stable configuration of the Au/Pd bimetallic alloy.22
Figure 1a, c and Table 1 present the process of Pd!Auswapping in both clean Pd!Au(111) and Pd!Au(100)surfaces, the driving force and the activation energy barrier(Eb), and the approximate rate of swapping is calculated at 300
K with harmonic-transition state theory, assuming a standardprefactor of ! = 1012 s!1. A Pd atom thermodynamically prefersthe subsurface layer, as compared with the surface layer. ThePd!Au swapping process, the barrier, and the rate have beencalculated in the direction where the Pd!Au swappingstabilizes the system, swapping a surface Pd atom with asubsurface Au atom.In both cases, an adjacent surface Au atom to the Pd atom
moves onto the surface to become an adatom, to make space (avacancy) for the Pd!Au swapping process. Adatom/vacancyformation is a high-energy process, and the energy barrier forPd!Au swapping is also (comparably) high. The swappingbarrier of the more open Pd!Au(100) surface is onlysomewhat lower than that of the close packed Pd!Au(111)(1.31 and 1.53 eV, respectively, as listed in Table 1). The lowcalculated rates confirm that, in the case of clean surfaces, Pdatoms would be pinned to their original positions at lowtemperature; even the Pd subsurface segregation is thermody-namically favorable (Table 1). Measurable swapping rates of 10s!1 are achievable at 650 K for Pd!Au(111) and 470 K in Pd!Au(100), respectively.
Figure 1. Pd!Au swapping process and relative energy of swapping intermediates in clean Pd!Au surfaces: (a) Pd!Au(111), (b) Pd!Au(111)!CO, (c) Pd!Au(100), and (d) Pd!Au(100)!CO. The relative energy of the intermediates was calculated relative to the unstable position of the Pdatom. The Pd!Au swapping proceeds from left to right, stabilizing the system. Au atoms involved in the swapping process are colored in light greenand pink.
Table 1. Pd!Au Swapping Energy Barrier (Eb) and theApproximate Rate of Pd!Au Swapping Calculated at 300 Kwith Harmonic-Transition State Theory, Assuming aStandard Prefactor of ! = 1012 s!1a
Pd!Au(111)Pd!
Au(111)!COPd!Au(111)-
VacPd!Au(111)-Vac!CO
Eb (eV) 1.53 1.56 0.73 0.81rate(s!1)
1.98 ! 10!14 6.21 ! 10!15 5.45 ! 10!1 2.47 ! 10!2
Pd!Au(100)
Pd!Au(100)!
COPd!
Au(100)-VacPd!Au(100)-Vac!CO
Eb (eV) 1.31 0.76 0.42 0.51rate(s!1)
9.84 ! 10!11 1.71 ! 10!1 8.80 ! 104 2.71 ! 103
aEb and the rate of Pd!Au swapping were calculated to the directionthat lowers the energy of the system.
We found that two Pd atoms in the surface layer repel andfavor being separated from each other. The formation energy ofa Pd!Pd dimer from two separated Pd atoms is 0.1 eV,meaning that Pd cluster formation is unfavorable. Configura-tional entropy would additionally destabilize the Pd!Pd dimerat low Pd surface concentrations. The swapping barrier of thesecond Pd atom in the presence of a preswapped Pd atom inthe surface layer would be higher than the barrier of the firstswapping.On the other hand, strong CO binding to Pd stabilizes the
Pd!CO* complex in the surface layer and supplies a drivingforce for the Pd surface segregation. (Figure 1b, d and Table 1).CO lowers the Eb of Pd surface segregation, especially in thecase of Pd!Au(100). Presumably, relatively stronger CObinding on the Au(100) surface (!0.63 eV, Au!Au bridgeposition) than the Au(111) surface (!0.25 eV, 3-fold hollowposition) lowers the energy of swapping intermediates in thePd!Au(100)!CO complex. However, the Pd!Au swappingstill requires an adatom formation in both cases, so the Eb is stillhigh (Table 1). Although CO exchanges the stable location ofPd, the Pd!Au swapping is a rare event, even in the presence ofCO.3-2. Energetics of Pd!Au Swapping in Defected
Surfaces. In their polarization!modulation infrared reflectionadsorption spectroscopy (PM-IRRAS) study on the well-annealed and freshly ion-sputtered Au-overlayered Pd(100),Goodman and co-workers reported that CO-induced Pdsurface segregation is more prominent in the freshly ion-sputtered Au/Pd(100) surface.15 Contiguous Pd atoms, a resultof high Pd surface concentration, were observed in the freshlysputtered specimen, even when it was exposed to very low COpartial pressure (1 ! 10!6 Torr). On the other hand, higher COpartial pressure is required for the formation of contiguous Pdatoms in the well-annealed specimen.15 This finding suggests acritical effect of surface roughness or defects for the dynamicsof the Pd!Au swapping. Moreover, recent experimentalfindings on the structure of Au/Pd bimetallic NPs confirmthe presence of surface vacancies in small NPs, as well.23,24
HRTEM studies by Xu et al. showed that vacancies appear inthe surface layer of Pd!Au NPs.24 Meija-Rosales et al.experimentally observed surface vacancies in Au!Pd NPs andconfirmed the structure by molecular dynamics simulations.23
These findings suggest the consideration of a defect on the Pd!Au swapping processes by introducing a Au surface vacancy.Figure 2a, c shows that a Au vacancy facilitates the Pd!Au
swapping pathway without adatom formation or surfacedistortion. A Au vacancy in the surface layer of the Pd!Au(111)-Vac and Pd!Au(100)-Vac, therefore, lowers theswapping energy barriers (Table 1). Pd penetration from thesurface layer to the subsurface layer is the rate-determining stepdue to the relatively low energy of the intermediate structure:the Au/Pd surface with a subsurface Au vacancy.Figure 2b, d shows that the preferential CO adsorption on
Pd again stabilizes the Pd!CO* complex in the surface of thedefected Pd!Au(111)-Vac!CO and Pd!Au(100)-Vac!CO,leading to surface segregation of Pd. The presence of CO againsupplies a driving force for Pd surface segregation but does notaccelerate the Pd!Au swapping (Table 1).In the case of the Pd!Au(111)-Vac systems (Figure 2a, b)
DFT predicts a structure with a subsurface Au vacancy and a Pdin the surface layer as a more stable structure than the final statewith a surface vacancy. A similar result was acquired for theclean Au(111) surface, as well; the result is insensitive to the
calculation parameters, including k-points grid, energy cutoff,and slab thickness. This is a somewhat surprising result becausethe subsurface vacancy generates more dangling bonds than thesurface oxygen vacancy. However, because the formation of thesurface Pd!CO* is insensitive to the location of the Auvacancy, the main conclusions reached here are not affected.Our findings confirm that, under CO oxidation conditions,
CO supplies a driving force for the preferential surfacesegregation of CO-philic Pd atoms in the Au/Pd bimetallicalloys so that the local geometry of the Au/Pd bimetallic alloyscould be different from their thermodynamically most stablestructure. Although CO molecules affect the Pd!Au swappingbarrier, we find that the vacancy critically accelerates the Pd!Au swapping (Table 1).
3-3. Multiadsorption of CO on Pd Atoms andSubsequent Pd Clustering. Goodman and co-workersreported that the Au/Pd(100) surface alloy catalyzes COoxidation as contiguous Pd atoms in the surface layerdissociates O2 molecules.14,15 As experimental evidence, theyresolved PM-IRRAS data acquired at 100 K and reported IRpeaks of the bridging CO species, Pd!CO*!Pd, at 1999, 1976,and 1908 cm!1. These peaks are located below 2000 cm!1,whereas the peak that corresponds to Pd!CO* lies at 2085cm!1 (Figure 3a).DFT-calculated IR frequencies of the Pd!CO*!Pd were
found at 1910, 1904, and 1902 cm!1, which are in goodagreement with the experimental value of 1908 cm!1 (Figure3d, e). We found that the experimental peak at 1999 cm!1 iscoming from the harmonics between adjacent two Pd!CO*species (see Figure 3c). The experimentally reported IR peak at1976 cm!1 is likely due to bridge CO molecules bound to Pdatom clusters. We found a frequency at 1930 cm!1 (Figure 3e)in a Pd6 cluster model. Additional subsurface Pd atoms wouldshift these values to higher energies. DFT-calculated IRfrequencies of weakly bound Pd!CO*!Au species werefound at 1893 and 1891 cm!1 (Figure 3b, d, and e). These
Figure 2. Pd!Au swapping process and relative energy of swappingintermediates in defected Pd!Au surfaces: (a) Pd!Au(111)-Vac, (b)Pd!Au(111)-Vac!CO, (c) Pd!Au(100)-Vac, and (d) Pd!Au(100)-Vac!CO. Relative energy of intermediates was calculated relative tothe unstable position of the Pd atom. The Pd!Au swapping proceedsfrom left to right, stabilizing the system. A Au atom involved in theswapping process is colored in light green.
modes, however, would disappear as the Pd!CO* and Pd!CO*!Pd become a dominant species, at increased Pdconcentrations.Given the well-known CO binding nature on CO-philic
metal surfaces (single CO adsorption on a metal atom), COstripping has been used to determine the surface coverage ofCO-philic metal elements of bimetallic systems.25 Galhenage etal. found that the surface concentration of Co, Ni, or Pt of Au-based bimetallic clusters supported on TiO2 estimated by COtemperature-programmed desorption is greater than that from alow-energy ion scattering experiment.25 We postulate that thepresence of the bridge-bound CO is attributed to theoverestimated surface concentration of the CO-philic elementestimated by CO temperature-programmed desorption.Soto-Verdugo and Metiu showed that the repulsive force
between adjacent Pd!CO*’s is low; Supporting InformationFigure S1 confirms their finding.16 Because the binding energyof the bridging CO molecule (Pd!CO*!Pd) is higher than theon-top Pd!CO* molecule, as Pd atoms segregate to the surface(see Figure 4), these bridge-bound CO molecules wouldpromote Pd clustering.
3-4. CO Oxidation by Pd Motifs. To provide insight intothe reactive species, we studied CO oxidation by several COspecies at the surface of Pd!Au(100): (1) Pd monomer with asingle on-top CO, Pd!CO* (M1), (2) Pd monomer with twobridging CO’s, Pd!CO*!Au (M2), (3) Pd dimer with two on-
top CO’s, Pd!CO* (D1), and (4) Pd dimer with threebridging CO’s, one bridging Pd!CO*!Pd, and two bridgingPd!CO*!Au’s (D2). Refer to Figure 5 for the detailedgeometry of these models. We found, however, that thesespecies cannot bind O2 strongly enough to catalyze COoxidation. Under CO oxidation conditions, M1 would prefer tobind an additional CO molecule (Eb = !0.96 eV), forming theM2 structure rather than binding an O2 molecule with a lower(!0.44 eV) binding energy. In the case of M2, the available Pdsites are already saturated by CO molecules so thatcoadsorption of O2 with the two CO molecules (!0.27 eV)is weak. Note that the binding energy of O2 on M2 is lowerthan the entropic contributions to the free energy of O2desorption, !0.64 eV (the entropic contribution to the Gibbsfree energy of O2 desorption at the conventional operatingtemperature of CO oxidation is !0.64 eV at 298 K and 1 bar;the standard entropy of O2 at 298 K is 205.14 J mol!1 K!1),26,27
confirming that additional binding of O2 on M2 is not favorable(see the Supporting Information for details). D1 would preferto bind additional CO (Eb = !1.21 eV) rather than O2 (Eb =!0.69 eV). Weak O2 binding at D2 (Eb = !0.22 eV) also showsthat D2 is not a good catalyst geometry for CO oxidation. COoxidation by a catalyst that weakly binds O2 and cannot supplya reactive O* species usually follows the Langmuir!Hinshel-wood mechanism, requiring the association of coadsorbed CO*and O2*.
9,10,28!30 Results show that isolated Pd atoms ordimers cannot activate CO oxidation by the association ofcoadsorbed CO* and O2*.Table 2 shows that the binding preference of the Pd cluster
changes from CO to O2 as the Pd cluster is composed of morethan four Pd atoms; Pd4 and Pd6 clusters strongly bind O2.Because DFT calculations at the GGA level of theory havesystematic errors in the binding energy of O2 and CO, therelative binding energies (for example, ΔEad in Table 2) areexpected to be more accurate than the absolute values. Thequalitative trend in the value of ΔEad, changing from positive tonegative at Pd4, leads us to conclude that the binding of O2 isfavored over CO in Pd clusters larger than Pd4. Moreover,
Figure 3. DFT calculated IR frequencies of surface CO species: (a, b) Pd monomer, (c, d) Pd dimer, and (e) Pd cluster consisting of six Pd atoms.Values in parentheses show experimental IR data.14,15
Figure 4. Strong binding of Pd!CO*!Pd on the Pd6 cluster. Thestronger binding of the final three CO molecules supplies a drivingforce for Pd clustering.
Figure 5. Trends in competitive CO and O2 binding on the Pd monomer (a) and Pd dimer (b) in the Pd!Au(100) surface. Pd monomer and dimerprefer to bind CO molecules as much as possible (green arrows) rather than binding an O2 molecule with CO molecules (red arrows). Theassociative mechanism of CO oxidation,9,10,28 CO oxidation by coadsorbed O2 and CO is not the case of the Pd monomer and dimer.
clustering of Pd atoms lowers the activation energy of O2dissociation, leading to easier O2 dissociation and promotingsubsequent CO oxidation by highly reactive O* that oxidizesCO by the Langmuir!Hinshelwood mechanism or the Eley!Rideal mechanism (refer to Supporting Information Figure S2for detailed geometries of O2 adsorption and dissociation of Pd4and Pd6).
9,10 We postulate that although some CO moleculeson the Pd atoms of Pd clusters that drive Pd surface segregationwould be removed from the surface as a result of CO oxidation,Pd clusters would stay in the surface layer under CO oxidationcondition as a result of strong O2 binding to the Pd clusters.Because the CO oxidation in this system is catalyzed by Pdmotifs larger than Pd4, the overall CO oxidation reactivitywould converge to those of the pure Pd (100) surface.Given the stronger CO binding energy of Pd motifs smaller
than Pd4 (see Table 2), CO binding on a Pd atom initiallysupplies a driving force for Pd surface segregation. As the size ofPd motifs increase larger than Pd4, the strong oxygen bindingon Pd motifs larger than Pd4 would attribute to further Pdsurface segregation, as well.Because the Au!Pd catalyst is exposed to CO oxidation
conditions (a mixture of CO and O2), the reduction of the Pd!O* by CO would be very fast (by the Eley!Ridealmechanism). The PdO oxide islands, therefore, would not bestabilized (the life span of the oxide at the surface layer of Pd!Au alloy would be short). Even though ideal theoreticalcalculations could predict the formation of PdO islands inoxygen-rich conditions, it would not be the case of the real COoxidation conditions.
4. SUMMARYAccording to conventional computational catalyst designmethods, which have focused on ensemble and ligand effects,a low concentration of Pd in a Au alloy would not likely beregarded as a catalyst for CO oxidation because Au surfaceatoms cannot bind and dissociate O2. Herein, we suggest,however, that under CO oxidation conditions where the CO-induced Pd surface segregation occurs, the Pd!Au(100)bimetallic alloy becomes an effective CO oxidation catalyst.The strong CO-philic nature of Pd supplies a driving force forpreferential surface segregation of Pd atoms, and a Au vacancydramatically accelerates the Pd!Au swapping. This findingpredicts that an adsorption-induced surface segregation wouldbe more prominent in nanoparticles or rough surfaces, wheresurface atoms are less closely packed.Pd clusters composed of at least four Pd atoms are found to
be a reactive species for CO oxidation. Facile O2 dissociation byPd clusters is essential for high CO oxidation activity. Our
findings suggest that not only are geometric and electroniceffects important, but dynamical effects also have to beconsidered for bimetallic alloys or NPs whose componentsasymmetrically interact with reacting molecules.
! ASSOCIATED CONTENT*S Supporting InformationDetails about the DFT calculations of the free energy of O2binding and CO repulsion are provided. This material isavailable free of charge via the Internet at http://pubs.acs.org.
! AUTHOR INFORMATIONCorresponding Author*E-mail: [email protected], [email protected] Address†Center for Functional Nanomaterials, Brookhaven NationalLaboratory, Upton, New York, 11973NotesThe authors declare no competing financial interest.
! ACKNOWLEDGMENTSWe gratefully acknowledge support from the ChemicalSciences, Geosciences, and Biosciences Division, Office ofBasic Energy Sciences, Office of Science, U.S. Department ofEnergy (Contract: DE-FG02-13ER16428). Computing timewas provided by the National Energy Research ScientificComputing Center and the Texas Advanced ComputingCenter at the University of Texas at Austin.
! REFERENCES(1) Crooks, R. M.; Ye, H. C. J. Am. Chem. Soc. 2007, 129, 3627!3633.(2) Greeley, J.; Mavrikakis, M. Nat. Mater. 2004, 3, 810!815.(3) Hammer, B.; Norskov, J. K. Adv. Catal. 2000, 45, 71!129.(4) Gross, A. Top. Catal. 2006, 37, 29!39.(5) Norskov, J. K.; Bligaard, T.; Rossmeisl, J.; Christensen, C. H. Nat.Chem. 2009, 1, 37!46.(6) Greeley, J.; Stephens, I. E. L.; Bondarenko, A. S.; Johansson, T.P.; Hansen, H. A.; Jaramillo, T. F.; Rossmeisl, J.; Chorkendorff, I.;Norskov, J. K. Nat. Chem. 2009, 1, 552!556.(7) Honkala, K.; Hellman, A.; Remediakis, I. N.; Logadottir, A.;Carlsson, A.; Dahl, S.; Christensen, C. H.; Norskov, J. K. Science 2005,307, 555!558.(8) Studt, F.; Abild-Pedersen, F.; Bligaard, T.; Sorensen, R. Z.;Christensen, C. H.; Norskov, J. K. Science 2008, 320, 1320!1322.(9) Kim, H. Y.; Han, S. S.; Ryu, J. H.; Lee, H. M. J. Phys. Chem. C2010, 114, 3156!3160.(10) Kim, H. Y.; Kim, D. H.; Ryu, J. H.; Lee, H. M. J. Phys. Chem. C2009, 113, 15559!15564.(11) Tao, F.; Grass, M. E.; Zhang, Y. W.; Butcher, D. R.; Renzas, J.R.; Liu, Z.; Chung, J. Y.; Mun, B. S.; Salmeron, M.; Somorjai, G. A.Science 2008, 322, 932!934.(12) Tao, F.; Grass, M. E.; Zhang, Y. W.; Butcher, D. R.; Aksoy, F.;Aloni, S.; Altoe, V.; Alayoglu, S.; Renzas, J. R.; Tsung, C. K.; Zhu, Z.W.; Liu, Z.; Salmeron, M.; Somorjai, G. A. J. Am. Chem. Soc. 2010, 132,8697!8703.(13) Tenney, S. A.; Ratliff, J. S.; Roberts, C. C.; He, W.; Ammal, S.C.; Heyden, A.; Chen, D. A. J. Phys. Chem. C 2010, 114, 21652!21663.(14) Gao, F.; Wang, Y. L.; Goodman, D. W. J. Phys. Chem. C 2009,113, 14993!15000.(15) Gao, F.; Wang, Y. L.; Goodman, D. W. J. Am. Chem. Soc. 2009,131, 5734!5735.(16) Soto-Verdugo, V.; Metiu, H. Surf. Sci. 2007, 601, 5332!5339.(17) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15!50.
Table 2. Trends in CO Adsorption and O2 Adsorption andDissociation on Pdx Clusters
a
Pd1 Pd2 Pd4 Pd6EadCO (eV) !1.14 !1.17 !1.28 !1.30
EadO2 (eV) !0.77 !1.06 !1.57 !1.66
ΔEad = EadO2 ! Ead
CO 0.37 0.09 !0.29 !0.36O2 dissociation energy (eV) n/a n/a 0.04 !0.86Eb (eV) n/a n/a 0.28 0.16
aEadCO, Ead
O2, and Eb denotes average energy of CO adsorption, energy ofO2 adsorption, and O2 dissociation barrier, respectively. Pdx is a Pdcluster composed of x Pd atoms on the Pd!Au(100) surface. EadCO wascalculated with two, four, six, and nine bound CO molecules to Pd1,Pd2, Pd4, and Pd6, respectively.
(18) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,3865!3868.(19) Blochl, P. E. Phys. Rev. B 1994, 50, 17953.(20) Henkelman, G.; Jonsson, H. J. Chem. Phys. 2000, 113, 9978!9985.(21) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. J. Chem. Phys.2000, 113, 9901!9904.(22) Ruban, A. V.; Skriver, H. L.; Norskov, J. K. Phys. Rev. B 1999,59, 15990!16000.(23) Mejia-Rosales, S. J.; Fernandez-Navarro, C.; Perez-Tijerina, E.;Blom, D. A.; Allard, L. F.; Jose-Yacaman, M. J. Phys. Chem. C 2007,111, 1256!1260.(24) Xu, J.; White, T.; Li, P.; He, C. H.; Yu, J. G.; Yuan, W. K.; Han,Y. F. J. Am. Chem. Soc. 2010, 132, 10398!10406.(25) Galhenage, R. P.; Ammal, S. C.; Yan, H.; Duke, A. S.; Tenney, S.A.; Heyden, A. J. Phys. Chem. C 2012, 116, 24616!24629.(26) Atkins, P.; Paula, J. D. Physical Chemistry, 8th ed.; Oxford: NewYork, 2006.(27) Metiu, H. Physical Chemistry: Statistical Mechanics, 1st ed.;Taylor & Francis: New York, 2006.(28) Liu, Z.-P.; Hu, P.; Alavi, A. J. Am. Chem. Soc. 2002, 124, 14770!14779.(29) Falsig, H.; Hvolbaek, B.; Kristensen, I. S.; Jiang, T.; Bligaard, T.;Christensen, C. H.; Norskov, J. K. Angew. Chem., Int. Ed. 2008, 47,4835!4839.(30) Kim, H. Y.; Lee, H. M.; Henkelman, G. J. Am. Chem. Soc. 2012,134, 1560!1570.
